Helmet Violation Detection System Based on YOLO11 and EfficientNet for Real-Time Traffic Monitoring

Helmet usage is critical to motorcycle safety, especially in traffic-dense countries like Indonesia. The following research offers a real-time helmet violation detecting system with a blend of YOLO11’s object detection capability and classification potency of the EfficientNet-B0 deep CNN model. The YOLO11 model was trained on a proprietary dataset of 1,888 annotated images with annotated motorcyclists. Images of riders' head regions, after detecting motorcyclists, are classified for helmet use with a fine-tuned EfficientNet-B0 model. The system performed well with YOLO11 model obtaining 94.1% accuracy and 99.1% mAP@50. EfficientNet-B0 classifier also performed well with training and validation accuracy with reinforcement from preprocess mechanisms such as denoising and contrast stretching. The results of the experiments confirm that the system can process incoming video frame with steady state performance and can effectively detect helmet violations and hence can be a viable solution for traffic monitoring with surveillance cameras.